Most people meet a new AI tool the same way. They open a chat window and start typing. “Write me some marketing copy.” “Summarise this.” “Give me ideas for my business.” Something comes back, it’s fine, they tweak it, and a week later they couldn’t tell you what any of it changed.

That’s not an AI problem. It’s a strategy problem. You skipped the thinking that makes the tool useful, so you got a tool doing something vaguely in the right area.

I built PAST — Purpose, Audience, Scope, Tone — to be the thinking you do before you type. Four questions, thirty seconds. They turn a pile of scattered experiments into one focused piece of work you can actually judge. Here’s each one, with a solopreneur on the other end of it.

Purpose: what outcome are you after?

The wrong question is “how can I use AI?” The right question is “what needs to be different when this is done?”

Say you’re a freelance bookkeeper drowning in client onboarding. “Use AI for admin” is not a purpose. “Cut new-client intake from three hours to forty minutes so I can take on two more clients without adding hours” is. It names the outcome, the size of the win, and how you’d know you got there.

If you can’t finish the sentence “success means ___, measured by ___,” you’re not ready to prompt. You’re ready to think a bit longer.

Audience: who is this for?

Every AI output has someone on the receiving end, and it’s almost never you.

A prompt that asks for “a summary of this report” ignores the most important variable: who reads the summary. Your accountant and your least technical client need completely different things from the same source document. One wants the numbers and the caveats. The other wants three sentences and a clear next step.

Name the reader before you write the prompt. What do they already know? What will they do with this? What would make them trust it? A solopreneur writing a proposal for a cautious corporate buyer and one writing a quick note to a client of ten years are doing two different jobs, even when the underlying facts are identical.

Scope: how big is this slice?

Scope creep kills more small-business AI projects than anything technical. Every tool can do dozens of things, and the pull is to point it at all of them at once. Then nothing quite works and you can’t tell why.

Pick one slice. If you’re a solo marketer, don’t “automate your content.” Automate the first draft of your weekly newsletter, using only your own past issues as source material, in the format you already use. That’s a boundary you can build, test, and trust in an afternoon.

Deciding what’s out is the real work here:

  • In: the newsletter first draft.
  • Out: subject lines, images, the actual send, anything you haven’t done by hand enough times to judge.

Prove value in the narrow version. Widen only once it earns the room.

Tone: how should it feel?

This is the one people skip, and it’s the one that gives AI away. Generic AI output has a smell: the flawless grammar, the “impressive” and the “innovative,” the copy that could belong to any business in your field. Send that to a client and you’ve told them you didn’t really show up.

Tone is where your voice goes back in. If you write to clients like a person, all short sentences and dry specifics, say so in the prompt and hand it a real sample. Better still, treat the AI draft as scaffolding and put the specifics in yourself: the actual project, the real number, the thing only someone who did the work would know. The machine handles the structure. You handle the parts that sound like you.

The thirty seconds that saves the afternoon

None of this is complicated, and that’s the point. Purpose, Audience, Scope, Tone. You can run them in your head before you touch the keyboard.

The payoff isn’t a faster first draft. It’s fewer drafts. You get usable output on the first or second try instead of the tenth, because you decided what “usable” meant before you asked. A random AI experiment goes nowhere because nobody defined where “somewhere” was. PAST is you answering that up front, in the order that matters.

Start with the one job you keep meaning to hand to AI and haven’t. Run it through the four questions. If you’d like a second pair of eyes on your answers, get in touch.